AIM: To determine the stability of beliefs of patients with Type 2 diabetes about their diabetes over 3 years, following diagnosis. METHODS: Data were collected as part of a multicentre cluster randomized controlled trial of a 6-h self-management programme, across 207 general practices in the UK. Participants in the original trial were eligible for follow-up with biomedical data (HbA1c levels, blood pressure, weight, blood lipid levels) collected at the practice, and questionnaire data collected by postal distribution and return. Psychological outcome measures were depression (Hospital Anxiety and Depression Scale) and diabetes distress (Problem Areas in Diabetes scale). Illness beliefs were assessed using the Illness Perceptions Questionnaire-Revised and the Diabetes Illness Representations Questionnaire scales. RESULTS: At 3-year follow-up, all post-intervention differences in illness beliefs between the intervention and the control group remained significant, with perceptions of the duration of diabetes, seriousness of diabetes and perceived impact of diabetes unchanged over the course of the 3-year follow-up. The control group reported a greater understanding of diabetes during the follow-up, and the intervention group reported decreased responsibility for diabetes outcomes during the follow-up. After controlling for 4-month levels of distress and depression, the perceived impact of diabetes at 4 months remained a significant predictor of distress and depression at 3-year follow-up. CONCLUSIONS: Peoples' beliefs about diabetes are formed quickly after diagnosis, and thereafter seem to be relatively stable over extended follow-up. These early illness beliefs are predictive of later psychological distress, and emphasize the importance of initial context and provision of diabetes care in shaping participants' future well-being.